Genome-wide association studies (GWAS) have been successful in identifying common genetic variants contributing to disease risk. However, nearly all of these studies have been conducted in populations of European ancestry. It is important to include other populations, because GWAS in Europeans are unlikely to detect risk variants that are common only in non-European populations. In the United States, the majority of individuals of non-European ancestry belong to admixed populations (e.g. African Americans or Latinos) that inherit ancestry from more than one continental population. Existing GWAS methods for admixed populations are inadequate, because they do not incorporate both SNP association and admixture association signals. Thus, if existing methods are applied, analyses will not be fully powered and important variants will be missed. For diseases with known population differences-such as cardiovascular disease in African Americans and asthma in Latinos-the need to develop methods that combine these signals is particularly pressing, because admixture association signals are likely to be particularly important. Here, we propose to develop a complete set of methods and software to combine SNP and admixture association signals in GWAS in admixed populations, while addressing questions such as imputation and choosing SNPs for replication. Our goal is to make fully powered association studies in populations of mixed ancestry as practical as studies in populations of homogeneous ancestry. In addition to African Americans, we will also develop methods for complex admixed populations (e.g. Latinos) that inherit ancestry from three or more continental populations, and for related individuals from admixed populations. Our methods research will be driven by empirical data, including over 10,000 African American samples and 2,400 Latino samples that will be genome-scanned by the CARe consortium, the Jackson Heart Study, and the Multiethnic Cohort Study. Our work will be applicable not only to GWAS in admixed populations, but also to meta-analyses of European and admixed populations, as well as future resequencing-based studies.